Allison Lee
East Brunswick High School
East Brunswick, New Jersey
MergeFinder: A Deep Learning Model for Galaxy Merger Identification
Allison found a way to use machine learning to identify places in the observable universe where galaxies collide.
View Poster
Allison Lee, 17, of East Brunswick, developed a convolutional neural network (CNN) to search space for galactic collisions for her Regeneron Science Talent Search space science project. Cosmologists rely on vast troves of galaxy images to study the universe. Processing and correcting data at that scale is a mathematical challenge.
In her project, Allison found a bias in the datasets that cosmologists typically use. So, she created a CNN called MergeFinder to correct that bias. Allison trained MergeFinder on simulations of galaxies merging and invented a new machine-learning technique to sort galactic data more efficiently. Her technique, called sequential classification, uses key features to sort data into categories and subcategories before recombining it for a final classification round. When used on the flawed datasets, MergeFinder found 2,208 new potential merger sites. These sites could give us clues about the large-scale structure of the universe and its origins.
Allison is the child of Eileen Liang and Yi-Horng Lee. She attends East Brunswick High School, where she produces and records the school’s daily announcements for the video newsmakers club. Allison also responds to local medical emergencies as part of the all-volunteer Milltown Rescue Squad.
Beyond the Project
Over the summer, Allison worked as a research assistant at Princeton University. While there, she developed software to improve the quality of electron microscopy images.
FUN FACTS: Allison is concertmaster of the New Jersey Symphony Youth Orchestra and one of four youth musicians selected for the Anne Lieberson Ensemble.
